2 * Copyright (c) 2015 Paul B Mahol
4 * This file is part of FFmpeg.
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 #include "libavutil/attributes.h"
22 #include "libavutil/cpu.h"
23 #include "libavutil/x86/cpu.h"
24 #include "libavfilter/blend.h"
26 #define BLEND_FUNC(name, opt) \
27 void ff_blend_##name##_##opt(const uint8_t *top, ptrdiff_t top_linesize, \
28 const uint8_t *bottom, ptrdiff_t bottom_linesize, \
29 uint8_t *dst, ptrdiff_t dst_linesize, \
30 ptrdiff_t width, ptrdiff_t height, \
31 struct FilterParams *param, double *values, int starty);
33 BLEND_FUNC(addition, sse2)
34 BLEND_FUNC(grainmerge, sse2)
35 BLEND_FUNC(average, sse2)
37 BLEND_FUNC(darken, sse2)
38 BLEND_FUNC(grainextract, sse2)
39 BLEND_FUNC(multiply, sse2)
40 BLEND_FUNC(screen, sse2)
41 BLEND_FUNC(hardmix, sse2)
42 BLEND_FUNC(divide, sse2)
43 BLEND_FUNC(lighten, sse2)
45 BLEND_FUNC(phoenix, sse2)
46 BLEND_FUNC(subtract, sse2)
48 BLEND_FUNC(difference, sse2)
49 BLEND_FUNC(difference, ssse3)
50 BLEND_FUNC(extremity, sse2)
51 BLEND_FUNC(extremity, ssse3)
52 BLEND_FUNC(negation, sse2)
53 BLEND_FUNC(negation, ssse3)
55 av_cold void ff_blend_init_x86(FilterParams *param, int is_16bit)
57 int cpu_flags = av_get_cpu_flags();
59 if (EXTERNAL_SSE2(cpu_flags) && param->opacity == 1 && !is_16bit) {
60 switch (param->mode) {
61 case BLEND_ADDITION: param->blend = ff_blend_addition_sse2; break;
62 case BLEND_GRAINMERGE: param->blend = ff_blend_grainmerge_sse2; break;
63 case BLEND_AND: param->blend = ff_blend_and_sse2; break;
64 case BLEND_AVERAGE: param->blend = ff_blend_average_sse2; break;
65 case BLEND_DARKEN: param->blend = ff_blend_darken_sse2; break;
66 case BLEND_GRAINEXTRACT: param->blend = ff_blend_grainextract_sse2; break;
67 case BLEND_DIVIDE: param->blend = ff_blend_divide_sse2; break;
68 case BLEND_HARDMIX: param->blend = ff_blend_hardmix_sse2; break;
69 case BLEND_LIGHTEN: param->blend = ff_blend_lighten_sse2; break;
70 case BLEND_MULTIPLY: param->blend = ff_blend_multiply_sse2; break;
71 case BLEND_OR: param->blend = ff_blend_or_sse2; break;
72 case BLEND_PHOENIX: param->blend = ff_blend_phoenix_sse2; break;
73 case BLEND_SCREEN: param->blend = ff_blend_screen_sse2; break;
74 case BLEND_SUBTRACT: param->blend = ff_blend_subtract_sse2; break;
75 case BLEND_XOR: param->blend = ff_blend_xor_sse2; break;
76 case BLEND_DIFFERENCE: param->blend = ff_blend_difference_sse2; break;
77 case BLEND_EXTREMITY: param->blend = ff_blend_extremity_sse2; break;
78 case BLEND_NEGATION: param->blend = ff_blend_negation_sse2; break;
81 if (EXTERNAL_SSSE3(cpu_flags) && param->opacity == 1 && !is_16bit) {
82 switch (param->mode) {
83 case BLEND_DIFFERENCE: param->blend = ff_blend_difference_ssse3; break;
84 case BLEND_EXTREMITY: param->blend = ff_blend_extremity_ssse3; break;
85 case BLEND_NEGATION: param->blend = ff_blend_negation_ssse3; break;